Colorizing Grayscale CT Images of Human Lung Using Deep Learning
MetadataShow full metadata
The COVID-19 epidemic has broken out globally and will influence human healthy history perpetually. We notice that Computed Tomography (CT) images are grayscale ones from the point of view of digital signal visualization. Therefore, we contemplate whether the automatically rendering colours for the CT lung images via deep neural networks will contribute to diagnosing any diseases for medics. At present, our motivation for image colorization is inspired by the advancement of associated techniques, such as machine learning and artificial intelligence (AI), especially video analogies and transfer learning in the domain of deep learning. In this thesis, we experimented with two deep learning networks in completely distinct orientations for implementing the most reliable outcomes for colorizing the CT lung images: VGG-19 and ResNet based on exemplar colorization full-automatic colorization, respectively. For hybrid colorization, we select appropriate reference images so as to combine the style and content of the representations to colorize the target CT lung grayscale images. The colours of meat resemble those of human lungs, so the images of fresh pork, lamb, beef, and even rotten meat (for infected lungs) are collected for the hybrid colorization model. Moreover, three sets of training data consisting of painting and meat images are analysed to extract the per-pixel erudition for colorizing the greyscale CT lung images for the fully automatic approach. Pertaining to the results, we consider numerous methods (human visual analysis, PSNR, and SSIM) to evaluate the proposed deep neural network models. Compared with other techniques of colorizing CT lung images, the results of rendering the CT lung images by using deep learning are significantly genuine. The adoption of deep learning is a striking and adventurous endeavours.